WhatsApp Help Pages Leak Meta Chatbot Instructions Including ‘What Not to Say’

Original headline: WhatsApp Help Pages Leaked Hidden Instructions for a Meta Chatbot

GAI Press Editors · Tuesday, October 20, 2026 · AI summary

Approximately 80 pages of WhatsApp’s help documentation were found to contain internal instructions for Meta’s chatbot, including a list of prohibited phrases such as "WHAT NOT TO SAY." Meta said the exposure was caused by a bug.

The leak reveals Meta’s internal content controls, raising concerns over transparency and potential misuse.

Research angleHow does the public disclosure of internal AI chatbot guidelines affect user trust? — Trust Theory

Read the original at WIRED ↗Oct 20, 01:00 KST

This story could become your next paper

HJITM works with researchers from study design through surveys, analysis and submission to international journals in generative AI and technology management. Bring us your research question.

Ask about research design →Free statistics tool ↗About HJITM

← 5. US Bars China's DeepSeek AI from Governm…7. Making the Internet More Friendly to AI … →

This item is an AI-generated summary of reporting by WIRED. Please check the original article for accuracy. Copyright in the original article belongs to WIRED.